JPH01228072A - System for recognizing external shape of linear pattern - Google Patents

Abstract

PURPOSE:To obtain a correct external shape by returning the uneven external shape to the linear information of a mean point, dividing data into two, obtaining respective gravity centers, and approximating the external shape line based on a straight line to pass through the gravity center. CONSTITUTION:A photographed original picture (a) is binarized, a picture (b) is obtained, the boundary point of the binarization is averaged at every 8 scanning lines, the coordinates of the mean point is calculated, and made into a mean point picture (c). The whole picture is divided into two, for a left half, an area h1 which is obtained by eliminating the area, whose Y-coordinate is (a) from maximum and minimum values, is made into a processing area, it is vertically divided into two by the gravity center of the whole picture, and by linking the gravity centers of an upper half and a lower half, the mean line can be obtained. Further, only the mean points existing the outside of the line is objected, the mean line is repeatedly obtained in the same way, and the last mean line is made into the external shape line. For the right half, the same processing is executed.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention relates to a method for recognizing the outline of a linear pattern, and more specifically, it is used to recognize a micro semiconductor component having a rectangular outline such as a semiconductor pellet as a lead, stem, or sub-pellet. The present invention relates to an improvement in a recognition method that can accurately recognize the outer shape of a component such as a semiconductor pellet or the outer shape of a pattern thereof when bonding to a substrate or the like.

[Prior Art] Conventionally, micro semiconductor components with rectangular external shapes, such as semiconductor pellets, are bonded to leads, stems, sub-pellets, substrates, etc., and made into products. In some cases, the light emitting direction may be slightly misaligned depending on the position where the lead, stem, sub-pellet, substrate, etc. are attached.

[Problem to be solved] This positional misalignment is especially a problem with laser diodes, but the pellets with the light-emitting part are minute,
Its outer shape has small irregularities and is not a smooth straight line. Moreover, there are variations in their external shapes. Therefore, when sub-pellets, etc. are bonded to a substrate, a positional shift occurs between the pellets and the sub-pellets, and even if the sub-pellets, etc. are soldered to the substrate, the light emitting point of the product may not be at the desired center position. First, there is a drawback that the permissible range may be exceeded, resulting in a decrease in yield.

This invention solves the problems of the prior art, and provides a linear pattern outline that allows accurate recognition of the outline even if there are variations in the outline or small irregularities in the outline. The purpose is to provide a recognition method.

[Means for Solving Problem 6] Means for the linear pattern outer shape recognition method of the present invention to achieve such an object is that in the outer shape recognition method for recognizing the outer shape of a linear pattern, the linear pattern Each point of the binarized original image data obtained by imaging the outline of , divide them into two groups according to that one line, find the first average line that passes through the center of gravity of the average points of each group, and calculate the
A second average line passing through the center of gravity of each of the two groups is determined for the remaining plurality of average points located in a direction outward of the linear pattern from the average line of the plurality of average points of each of the two groups, This second or third and subsequent average lines obtained through similar processing are made to correspond to one line of the linear pattern.

[Operation] In this way, by binarizing the original image data obtained by imaging, dividing the data into groups, and averaging each group, the outline of the unevenness can be reduced to line information of the average point. Therefore, by dividing these data into two and finding the center of gravity of each,
The outline can be approximated by two straight lines passing through the center of gravity, and by moving the straight line data corresponding to the approximated outline outward, line information closer to the outline can be obtained.

As a result, more accurate outer shape data can be obtained, and by determining the center line and the like from this outer shape, the parts can be positioned accurately. In addition to positioning processing, when it is necessary to obtain an accurate external shape, the external shape can be obtained by simple processing.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a flowchart of a process of an embodiment in which the linear pattern outer shape recognition method of the present invention is applied to bonding a laser diode pellet to a sublet, and FIGS. 2(a) to (h) are: An explanatory diagram of the screen data shown on the display screen on the CRT display to explain the process, Figure 3 is an external view of the main parts of the laser diode pellet bonding equipment, and Figure 4 is the sub-pellet and laser diode pellet. FIG. 3 is a plan view, a side view, and an explanatory diagram showing a state in which they are joined.

In FIG. 3, 20 is a laser diode pellet bonding apparatus, and 1 is a bonding station for the laser diode pellet. The bonding station 1 includes a heating table 2 and a positioning claw mechanism 3.
A sub-pellet imaging camera 4 is provided on the upper part of the heating table 2, and a laser diode pellet imaging camera 5 is provided on the lower side adjacent to the heating table 2.

The video signals from the sub-pellet imaging camera 4 and the laser diode pellet imaging camera 5 are input to an image processing device 6, where a predetermined external shape recognition process is performed, and the video is displayed on a CRT display 6a serving as an image display section. Ru.

The heating table 2 has a cartridge heater embedded therein, and the temperature during bonding is controlled by a temperature regulator. On this heating table 2 are sub-pellets 21 (see Fig. 4) conveyed by the pellet handling mechanism 7.
a)) is positioned and held. The pellet handling mechanism 7 moves on an X table 8 in the X direction provided adjacent to the heating table 2, and its arm 7a is movable in the Z direction. A suction chuck 7b for suctioning the sub-pellet 21 is attached to the tip side of the arm 7a. In addition, the sub-pellets 21 are set in each partitioned frame of the pallet 9a placed on the Y table 9.

The sub-pellet 21 transferred by the pellet handling mechanism 7 is placed on the heating table 2, and its positioning is as follows.
This is done by abutting the sub-pellet 21 against the corner guide 2a by the positioning claw mechanism 3. the result,
The center of the sub-pellet 21 is positioned at a predetermined position on the heating table 2.

Here, the positioning pawl mechanism 3 is composed of a pawl 3a having a right-angled valley groove and an advancing/retracting mechanism (not shown) having a motor etc. for moving the pawl 3a forward and backward. Both the positioning claw mechanism 3 and the heating table 2 are placed on an X-Y table 10 that moves slightly, and their positions can be finely adjusted in the X/Y directions.

As shown in the plan view and side view (right side) of FIG. 4(a), the sub-pellet 21 is a rectangular silicon substrate, and its size is about 700 to 1000 μm square. A bonding surface 21a to which a laser diode pellet 22 (see FIG. 2B) is bonded is formed in an island shape in contact with one side of the bonding surface 21a. The joint surface 21a is approximately 30
It is approximately 0 μm square and serves as a solder pattern.

Placed on the bonding surface 21a of this solder pattern is a laser diode pellet 22 shown in FIG.
As shown in the plan view and side view (right side), the central portion thereof has a slightly protruding platform shape, and a rectangular bonding surface of the gold pattern 22a is formed there. This gold pattern 22a is placed in contact with the bonding surface 21a,
The solder pattern is heated and melted, and the laser diode pellet 22 is bonded and fixed to the sub-pellet 21.

Note that the laser diode pellets 22 are stored in compartments on a pallet lla on the loader table 11 with the gold pattern 22a facing down. The laser diode pellet 22 is then handled by a pellet handling mechanism 12 having a YZ-θ table and placed on the joint surface 21a of the sub-pellet 21 on the heating table 2. The pellet handling mechanism 12 moves on the X-table 8, 12a is an arm of this pellet handling mechanism 12, and 12b is a suction chuck provided at the tip side for sucking pellets. , which is supported by a θ table so that it can rotate.

Here, the laser diode pellet 22 chucked by the pellet handling mechanism 12 moves linearly in the An image is taken by a laser diode pellet imaging camera 5 on the side, and the image data is sent to an image processing device 6.

The image processing device 6 uses this image data to recognize the outer shape of the gold pattern 22a of the laser diode pellet 22 and detect its center line. Then, the laser diode pellet 2 is placed on the center line of the joint surface 21a of the sub-pellet 21.
so that the center line of 2 is on the center line of sub-pellet 21 (
(See center lines 0 and 0' in FIGS. 4(a) and 4(b)).

That is, as shown in FIG. 4(C), the center line 0' of the gold pattern 22a of the laser diode pellet 22
The sub-pellet 21 is placed on the joint surface 22a and joined so that it lies on the center line 0 of the joint surface 22a of the sub-pellet 21, and the sub-pellet 21 is placed on a lead or stem and molded to become a product. In this case, the position of the emitted light beam can be set to a predetermined position to reduce variations in the position of the emitted light beam, thereby improving the yield of the product.

High accuracy is required in the positional relationship between the laser diode pellet 22 and the sub-pellet 21, for example,
The accuracy is up to about ±25 μm with respect to the position of the sub-pellet 21 in the Y direction (see (C) in FIG. 4), and up to about ±10 as the mounting rotation angle θ.

In order to perform highly accurate positioning in this way, the heating table 2
Images of the sub-pellet 21 and the laser diode pellet 22 are taken by the sub-pellet imaging camera 4 and the laser diode pellet imaging camera 5 arranged above the sub-pellet 21 and the laser diode pellet 22, and their positioning is controlled by the image processing device θ.

In this case, the sub-pellet 21 in particular is relatively large in size compared to the laser diode pellet 22, and since it is positioned in abutment, it is difficult to accurately determine the position of its center line 0 through normal image processing. However, since the laser diode pellet 21 is small and has variations in external shape, it is difficult to accurately position the sub-pellet 21 by normal handling operations. Therefore, it is necessary to determine the center line 0' by grasping the accurate outer shape.

Therefore, a pattern recognition process for the outer shape of the gold pattern 21a of the sub-pellet 21 is performed in the image processing bag rIt6 according to the procedure shown in FIG.

In addition, in the image processing device 6, pixels (512X
512), and the following processing is performed on the pixel data stored there, but to simplify the explanation, this pixel data will be replaced with data on the screen display. do.

Figure 2 (a) showing the image on the CRT display ea.
As shown in FIG. 1, first, in step (2) in FIG. 1, the original image data of the laser diode pellet 22 is collected, and in step (2), an appropriate threshold is applied and these are binarized to obtain image pattern data of the gold pattern 22a. (see Figure 2(b)).

Next, in step ■,
As Y coordinate data (corresponding to the x and Y coordinates of the pixel position), the horizontal scanning of the image is grouped into groups of 8 lines, and the coordinate position of the average point (center of gravity) is calculated from the coordinate position of each point in these groups. Calculate and use it as average point image data. FIG. 2C shows the display state on the CRT display 6a at this time.

Next, in step (2), the average point image data is divided into two from the center of the screen in the X coordinate direction, and an average point group corresponding to the left half of the screen and an average point group corresponding to the right half are divided. Divide into first and second average score groups (see (d) in Figure 2).

In the next step (2), the outline of the gold pattern 22a of the laser diode pellet is first determined from the left side. That is, for each average point, the coordinate positions of the minimum and maximum Y coordinates are determined (see average point F and average point G in FIG. 2(d)).

In the next step ■, input the distance (number of pixels) a in the Y direction, and calculate the average point that is a distance a above the maximum point G in the Y direction and the distance in the Y direction from the minimum point F. a
Select a group h, whose average scores are within a range (inclusive) with the average score below. Here, the distance a is used to select a group of average points corresponding to the line in the Y direction, thereby eliminating unnecessary average point data. That is, here, the data corresponding to the line in the X direction is unnecessary and is therefore excluded. Therefore, the distance a is the first Y when viewed from the maximum point G or minimum point F.
The value may be equal to or greater than the larger distance to the corner point of the direction.

In the next step (2), the average point group h/ is divided into two groups approximately from the center in the Y direction. Then, in the next step (2), the centroids of these two groups of average points are determined, and an average line A connecting the two centroids is obtained (see line A in (e) of FIG. 2).

In the next step (2), the remaining average points outside the average line A in the image pattern of the gold pattern 22a are taken up for each group, and in step [phase], the remaining average points of the two groups of average points are taken up in the same way as above. Find each center of gravity and obtain an average line connecting these two centers of gravity (see Figure 2).
(See line B in e).

In this way, the average line is obtained two or more times (if it is three times or more, repeat steps ■ to step [phase]), and the average line finally obtained in step ■ is stored in the memory as the left contour line in the Y direction. do.

In the next step [phase], it is determined whether there is any remaining group among the two groups formed in step (2). At first, the YES condition is met, so return to step ■,
The remaining group on the right side is processed from step ■ to step [phase]. The process corresponding to this case is shown in FIG. 2(f), where the range of the group of average points is h2, and the average line C is the first line,
The average line is the second line.

In this way, when the left and right contour lines in the Y direction are obtained and stored in the memory, there are no remaining two groups, so in the step [phase], N
When the o condition is satisfied, the process moves to the next step [phase], and in step [phase], it is determined whether or not the determination of the outline in the X direction and the Y direction has been completed. At first, the outline in the Y direction is determined, and the X direction has not been determined yet, so the No condition is set here, and in the next step ■, the coordinates in the X direction and the coordinates in the Y direction are swapped, and the Return to ■. As a result, from now on, the outline in the X direction is obtained by the processing of steps ① to [phase]. h in Figure 2 (g)
a*11tt indicates each average point group in the X direction, and the distance S corresponds to the distance a in the Y direction, which is the average point K.

L, M, and N are its maximum point and its minimum point.

When the contour lines in the Y direction and the X direction are determined in this way, the YES condition is satisfied in the previous step @, and the process moves to the next process of calculating the center line.

From the four outlines thus obtained, each intersection of the gold pattern 22 is 15°, 16.17°, as shown in FIG. 2(h).
18, and if this external shape relationship is not within the allowable range for placement and positioning on the sub-pellet 21, the pellet handling mechanism 12 returns to its original position, chucks the laser diode pellet 22 again, and performs the process of re-recognizing it. I will do it. Note that, as shown in FIG. 2(h), when the outline has an inclination of θ11θ2 that exceeds the allowable range, the θ table of the handling mechanism 12 is rotated, and the suction chuck 12b is rotated to correct the inclination. .

When the alignment is within the allowable range, the sub-pellet 22
The center line of the joint surface 22a of the sub-pellet 22 and the gold pattern 22a are determined from the external shape information of the joint surface 22a of the sub-pellet 22 and the coordinate positions of the respective intersection points 15.1B and 17.18 of the gold pattern 22a.
The amount of deviation between the center line and the side P on the light emitting surface side ((
C)

Calculate the displacement coordinate amount in the Y direction and its rotational coordinate θ),
The sub-pellet 21 is made by the pellet handling mechanism 12.
The laser diode pellet 22 is placed on the bonding surface 2ia of the laser diode pellet 22 with the gold pattern 22a of the laser diode pellet 22 aligned.

The placed laser diode pellet 22 is placed on the heating table 2
is joined to the sub-pellet 22 by being heated. At this time, in order to prevent positional deviation, the laser diode pellet 22 may be chucked and fixed by the suction chuck 12b.

As explained above, in the example, the average score is calculated for 8 points as one group, but it is sufficient to calculate the average score for a plurality of adjacent points, and the number is not limited to 8. . Furthermore, it goes without saying that a plurality of adjacent points may partially overlap or may be skipped.

Further, in the embodiment, an example is given in which a laser diode pellet is bonded to a sub-pellet, but the present invention is not limited to such a case, and is applicable to pattern recognition of an electronic component having a straight line shape. This method can be applied to the recognition of various external patterns such as the following.

Furthermore, the recognized pattern is not limited to being used for positioning.

[Effects of the Invention] As can be understood from the above explanation, in this invention, the original image data obtained by imaging is binarized, divided into groups, and averaged for each group. can be reduced to the line information of the average score.

Therefore, by dividing these data into two and finding the center of gravity of each, the outline can be approximated by two straight lines passing through the center of gravity, and the straight line data corresponding to this approximated outline can be moved outward. By doing this, it is possible to obtain line information that is closer to the outer shape.

As a result, more accurate outer shape data can be obtained, and by determining the center line and the like from this outer shape, the parts can be positioned accurately. In addition to positioning processing, when it is necessary to obtain an accurate external shape, the external shape can be obtained by simple processing.

[Brief explanation of the drawing]

FIG. 1 is a flowchart of a process of an embodiment in which the linear cane pattern external shape recognition method of the present invention is applied to bonding a laser diode pellet to a sublet, and FIGS. 2(a) to (h) are: An explanatory diagram of the screen data shown on the display screen on the CRT display to explain the process, Figure 3 is an external view of the main parts of the laser diode pellet bonding equipment, and Figure 4 is the sub-pellet and laser diode pellet. FIG. 3 is a plan view, a side view, and an explanatory diagram showing a state in which they are joined. DESCRIPTION OF SYMBOLS 1... Bonding station, 2... Heating table, 3... Positioning claw mechanism, 4... Sub-pellet imaging camera, 5... Laser diode pellet imaging camera, 6...
- Image processing device t, 7.12... Pellet handling mechanism, 20... Bonding station, 21.
... Sub-pellet, 21a... Bonding surface, 22... Laser diode pellet, 22a... Gold pattern. Patent applicant Hitachi Electronics Engineering Co., Ltd. Agent Patent attorney Kore Kajiyama

Claims (2)

[Claims] (1) In an outline recognition method that recognizes the outline of a linear pattern, each point of the binarized original image data obtained by imaging the outline of the linear pattern is divided into groups of a plurality of adjacent points, and each Find the average score of the group, select the plurality of average points along one line of the linear pattern, divide them into two along that one line, divide them into two groups, and calculate the average score of each group. The first average line passing through the center of gravity is determined, and the remaining average points of the plurality of average points of each of the two groups, which are located in the outer direction of the linear pattern from the average line, are calculated as follows. determining a second average line that passes through the center of gravity in each of the two groups, and making this second or third and subsequent average lines obtained through similar processing correspond to the one line of the linear pattern; An outline recognition method for linear patterns featuring the following. (2) The linear pattern is a bonding figure pattern of rectangular semiconductor pellets, and the average points of each divided group are developed into XY coordinates, and the coordinates of each average point forming the bonding figure pattern are is divided into two in the X direction (or Y direction), divided into the first two groups, and these two
Select the plurality of average points along one line in the Y direction (or The first average line passing through the barycentric coordinates of the average points of each group is determined, and the second average line of the plurality of average points of each of the second two groups is determined in the outward direction of the joined figure pattern. Find the second average line passing through the center of gravity of each of the second two groups for the plurality of average points at the positions, and calculate the third and subsequent average lines obtained by this second or similar processing. 2. The linear pattern outer shape recognition method according to claim 1, wherein the linear pattern outer shape recognition method is made to correspond to the one line.